In this study we developed a network biology approach to find putative novel switches in the cAMP-PKA signaling pathway in platelets. The network analyses were automated in R and can be adapted to include different proteins or an entirely different dataset. Our method can be used to repurpose existing datasets and provide a coherent overview of mechanisms involved to predict novel connections, by visually integrating multiple datasets. Here, this method was utilized to distil proteins from an existing phosphoproteomics dataset containing both activation and inhibition of platelets [15, 16]. Proteins not related to phosphorylation were inherently not included in the original phosphoproteomics datasets, as well as proteins excluded by our stringent selection, e.g., effectors downstream PKA after only ADP stimulation. However, such candidates might still be included through the addition of interactors using tools like the StringApp. Furthermore, although cAMP-PKA signaling is a major platelet inhibitory pathway, other pathways such as NO-cGMP/PKG signaling inhibit platelet activation. An advantage of our workflow is that it is designed in such a way that it could be readily adapted to process an alternate data selection or an entirely different dataset. Following our rationale, we were able to extract a subset of 30 proteins from this dataset, i.e., those proteins displaying both altered phosphorylation status upon sequential treatment of platelets with ADP plus iloprost and altered phosphorylation by iloprost treatment only. We postulate that the activity of these 30 proteins can be modified after ADP-mediated platelet activation and subsequently remodified after platelet inhibition with iloprost.
GO enrichment analysis revealed that multiple biological processes related to vesicle secretion, regulation of small GTPases and regulation of cyclin-dependent protein kinase activity were significantly overrepresented in our obtained list of 30 proteins. These processes are known to be important in platelet physiology. Platelets influence hemostasis by secreting granules and communicate with their environment by producing extracellular vesicles. Small GTPases play a pivotal in the control of vesicle trafficking and platelet aggregation or thrombus formation via αIIbβ3. Network analysis showed several clusters corresponding to these enriched biological processes to be well defined. We will discuss regulated proteins that are detected/detectable in the platelet proteome (27 out of 30, Fig. 4) and have a high GPS score for PKA according to the cut-off of the GPS tool, as our hypothesis is that downstream of the cAMP-PKA pathway proteins or processes are present that can serve as a “switch” in platelet activation and inhibition.
The largest cluster in our analysis is related to vesicle-mediated transport. Interesting proteins in this cluster included FGA, STXBP5, STON2, SEC22B, LRMP, and VTI1B. The role of these 6 proteins in platelet function has well been described for fibrinogen alpha chain (FGA) and less well for the others, as fibrin is a major constituent of thrombi and essential to normal hemostasis [29]. Syntaxin-binding protein 5 (STXBP5) has contradicting functions in platelets and endothelial cells, it promotes granule secretion in platelets but inhibits exocytosis in endothelial cells [30]. A single-nucleotide polymorphism in the STXBP5 locus has also been associated with a decreased thrombotic phenotype [31]. In this study STXBP5 is indicated as a putative PKA substrate, which is in line with other phosphoproteomics studies on resting platelets [32]. Stonin-2 (STON2) is involved in the endocytic machinery, synaptic vesicle recycling and clathrin coated vesicle uncoating [33]. To our knowledge Stonin-2’s function has not been described in platelets even though Stonin-2 mRNA expression is relatively high in platelets compared to other hematopoietic cells, making it an interesting target for future research. SEC22B is a membrane-resident trafficking protein that is required for ɑ-granule production in megakaryocytes and can interact with NBEAL2, which is associated with grey platelet syndrome. LRMP better known as inositol 1,4,5-triphosphate receptor associated 2 (IRAG2), interacts with the inositol 1,4,5-triphosposphate receptor in mice indicating a potential role in calcium homeostasis [34]. Platelet endocytosis/exocytosis is important in e.g., loading/release of α-granules. Studies have also implicated that integrin trafficking contributes to platelet activation and thrombosis by controlling their surface expression [35]. As more mechanistic studies show platelet endocytosis to be involved in platelet function [36], it is interesting to see a pronounced representation in our network analysis.
The second largest cluster of proteins was related to regulation of small GTPase mediated signal transduction and cell shape. This cluster included ARHGEF6, PPP1R14A, KALRN, MYO9B, and ABLIM3, all with a high PKA score. Not much is known about ABLIM3’s function in platelets. However, the other proteins have been studied extensively in relation to platelet function. ARHGEF6 has previously been reported to be a substrate for PKA and PKG and acts as a mediator in reducing Rac1-GTP levels leading to less outside-in platelet signaling [37]. Protein phosphatase 1 regulatory subunit 14A (PPP1R14A or CPI-17) is a phosphorylation-dependent inhibitor protein of myosin phosphatase. Phosphorylation at Thr-38 causes a conformational change that greatly increases its inhibitory potential [38]. Although phosphorylation happens mainly through PKC and Rho-associated protein kinase (ROCK), it has a potential PKA target consensus sequence, as indicated by the GPS 5.0 algorithm. Phosphorylation of PPP1R14A has been shown to regulate shape change in platelets trough calcium-independent signaling pathways. Platelet shape change is necessary for complete platelet activation. PPP1R14A could therefore be involved in a potential switch in platelet hyperreactivity as it can regulate myosin light chain phosphatase [39]. Another protein related to Rac/Rho protein signaling is kalirin (KALRN). In platelets, adenosine 5'-diphosphate-ribosylation factor 6 (ARF6) controls platelet spreading via integrin αIIbβ3 trafficking and can recruit KALRN to the plasma membrane leading to Rac activation [35, 40]. The ATP binding site within KALRN might be used as a starting point in the design of specific inhibitors. Inhibition of Rac1 by small molecules has been studied in the context of platelet secretion as a potential future antiplatelet drug [41]. MYO9B has recently been indicated to regulate RhoA activation though phosphorylation by PKA and PKG [42]. RhoA is a molecular switch controlled by GTPases. Platelet activation by ADP in RhoA knockout mice results in defective platelet function and unstable thrombus formation [43]. RAC1 and RHOB were central nodes interacting with most of these proteins. Rho GTPases are known key regulators of platelet cytoskeleton and platelet function and serve as molecular switches downstream of platelet surface receptors. It is therefore expected that these proteins emerge as potential candidate switches for platelet regulation from our selection.
The cluster related to cyclin-dependent kinases contained at least three interesting proteins all with significantly upregulated phosphorylation sites after ADP and iloprost treatment; CDK16/17 and Cyclin-Y (CCNY). Regulation of cyclin-dependent protein kinase activity might seem surprising at first sight as cyclin-dependent proteins belong to a class of proteins involved in cell cycle, transcription and mRNA processing and platelets do not have a nucleus and limited mRNA processing capacity [44]. However, recent focus on the role of non-coding RNA, nuclear receptors and post-transcriptional modifications in platelets reveal a more intricate transcriptional landscape than previously thought [45, 46]. In HEK293a cells, CDK16 is recruited to the plasma membrane and activated by CNNY. This process is regulated by Ser-153 and can be inhibited by PKA phosphorylation [47]. In line with these findings, we found Ser-153 to be a putative PKA phosphorylation site. Platelets lacking CCNY show decreased spreading and clot retraction, but increased adhesion to collagen [48]. Taken together, these new findings suggest that CCNY could play an important role in the outside-in signaling in platelets.
Finally, the smaller clusters and single nodes in our network also contain interesting proteins that are worth mentioning. PDE3A, CLDN5, GCSAML, MTSS1, ARPP19, MACF1, TBC1D23, UBE20, GAS2L1, PHKB, SCAMP3, CNST, RGCC, SLAIN2, and EIF3B. As expected, among the 30 proteins revealed by pathway analysis, we found proteins known to be involved in the cAMP-PKA signaling axis, e.g., PDE3A. While iloprost causes cAMP levels to go up, PDE3A hydrolyses cAMP. In diabetes mellitus, platelet hyperreactivity is believed to at least partly be caused by reduced platelet sensitivity to insulin, which leads to decreased endothelial PGI2 expression, increased P2Y12-mediated Gi activity and decreased platelet cAMP levels, thus leading to increased platelet activation [49, 50]. PKA-induced phosphorylation of PDE3A creates a negative feedback loop, indicating PDE3A being a key element in the cAMP/PKA pathway at least in initiation of platelet activation [51]. Although not directly connected to any other nodes in the protein network, it is foreseeable that PDE3A is returned from the analysis as an interesting target. In fact, the validity of the method presented in this study can be inferred by its identification of PDE3A.
As more large datasets are generated daily by innovative studies utilizing high-throughput methods, integration of these datasets to put them in a specific context are necessary. Finding novel players in platelet activation/inhibition can help us better understand pathologies where platelet hyperreactivity is prevalent. Further experimental studies using a more causal analysis approach can reveal what effect a phosphorylation event has on a protein's function [52]. Experimental knockout mice or using inhibitors in flow experiments for the aforementioned proteins are valid options to further validate the roles of these proteins in platelet hyperreactivity. Finally, this study shows the importance and benefit of data integration and visualization with existing tools and datasets to obtain a complete picture of complex molecular mechanisms involved.